Google rolls out Gemini 2.5 Pro for complex AI tasks

Google has introduced Gemini 2.5 Pro, its most advanced artificial intelligence model, designed to handle complex reasoning tasks with greater accuracy and efficiency.

The announcement was made in a blog post on Thursday, where the tech giant highlighted that Gemini 2.5 Pro outperforms previous models in reasoning, coding, and problem-solving.

The model has also debuted at the top of the LMArena leaderboard, which evaluates AI performance based on human preferences.

“Gemini 2.5 Pro is a state-of-the-art AI model capable of reasoning through its thoughts before responding, leading to improved accuracy,” Google stated.

According to the company, the new model excels in maths and science benchmarks, including GPQA and AIME 2025, without relying on test-time techniques like majority voting, which often increase computational costs.

It also recorded an 18.8 per cent score on Humanity’s Last Exam, a dataset designed by subject matter experts to test AI’s ability to tackle human-level knowledge challenges.

Google noted that Gemini 2.5 Pro is now available in Google AI Studio and the Gemini app for advanced users, with plans to launch it on Vertex AI soon.

The company also plans to introduce pricing options for scaled production use in the coming weeks.

The rollout of Gemini 2.5 Pro marks another step in Google’s push for AI models that incorporate better reasoning and decision-making capabilities, making them more useful in handling complex problem-solving, coding, and scientific research.

Meanwhile, OpenAI, the ChatGPT maker, directly challenges Google’s core business in search, while DeepSeek disrupts the AI ecosystem with cost-efficient models and open-source strategies, indirectly pressuring companies like Google and OpenAI to adapt.

OpenAI’s advancements in artificial intelligence, particularly its language models, have reshaped the industry, prompting Google to incorporate generative AI into its search products.

Meanwhile, China’s DeepSeek has emerged as a strong competitor with its cost-efficient, high-performing AI models like the R1 and the enhanced V3-0324.

These models have intensified competition with U.S. tech giants such as OpenAI and Google by delivering comparable capabilities at a fraction of the cost.